I have my C: drive mounted as /c and my D: drive mounted as /d. If I run this command:
rsync -v -a -e ssh localhost:/c/cygwin/ /d/cygwin/ If you don't have a D: drive or some other separate drive you can use as the target directory for this test (using something underneath /c is a bad idea, because you run the risk of rsync trying to copy the directory hierarchy that it's creating), you can do the test remotely from a separate box, i.e., "rsync -v -a -e ssh "remote-host:/c/cygwin/ /some/local/directory/" -- the failure described below still occurs. Rsync reliably hangs part of the way through the transfer. When I run "ps" after ^C'ing the hung rsync, I see: PID PPID PGID WINPID TTY UID STIME COMMAND 223 1 223 223 0 1000 09:08:06 /usr/bin/bash 261 1 261 261 1 1000 13:09:35 /usr/bin/bash 186 1 186 186 ? 1000 14:11:34 /d/scratch/jik/openssh_cvs-20010517/sshd O 191 186 186 191 ? 1000 14:12:24 /d/scratch/jik/openssh_cvs-20010517/sshd O 89 191 89 140 ? 1000 14:12:25 /usr/local/bin/rsync 205 223 205 197 0 1000 14:24:40 /usr/bin/ps Neither rsync nor sshd responds to kill -TERM -- I need to use "kill -9" to make them die. By stepping back through the OpenSSH anonymous CVS repository, I discovered that the first day on which this problem occurs is May 17. This is the first day in which Corinna's change to enable USE_PIPES appears. If I compile sshd.exe from the current OpenSSH anonymous CVS repository sources with the USE_PIPES line removed from configure.ac, the problem goes away. My understanding is that pipes behave somewhat poorly under Cygwin because pipes behave somewhat poorly under Windows :-). It would seem that Cygwin pipes don't behave reliably enough for OpenSSH to use them. I'd suggest removing USE_PIPES from OpenSSH for Cygwin. jik